Hi,
I’m using Gemini 3 flash to do batch data annotation. Each of my batches have around 200000 tokens (well below the 3M enqueued token limit for my tier – tier 1 see below). I do NOT have active jobs on the same model and the batch file is well below 2GB limit.
However, I keep receiving 429 RESOURCE_EXHAUSTED error for my request.
I urgently need to figure this out: my guess is when I cancel or delete an active job, those tokens are NEVER flushed from the enqueued token count.
Can someone from the Google team double check this?
